BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Michael Munn (Google) DTSTART:20241114T133000Z DTEND:20241114T143000Z DTSTAMP:20260423T005736Z UID:IMSRS/12 DESCRIPTION:Title: L everaging implicit bias to improve efficiencies in training and fine-tunin g ML models\nby Michael Munn (Google) as part of IUT Mathematics Resea rch Seminars (IMRS)\n\n\nAbstract\nIn classical statistical learning theor y\, the bias variance tradeoff describes the relationship between the comp lexity of a model and the accuracy of its predictions on new data. In shor t\, simpler models are preferable to more complex ones and\, in practice\, we often employ many techniques to control the model complexity. However\ , the best way to correctly measure the complexity of modern machine learn ing models remains an open question. In this talk\, we will discuss the no tion of geometric complexity and present some of our previous research whi ch aims to address this fundamental problem. We'll also discuss current an d future work which leverages this insight to devise strategies for more e fficient model pre-training and fine-tuning.\n LOCATION:https://researchseminars.org/talk/IMSRS/12/ END:VEVENT END:VCALENDAR